Dixon Woon - UAA Congress 2025

Associate Professor Dixon Woon is a consultant urologist at Eastern Health, Austin Health, and the Olivia Newton-John Cancer Wellness and Research Centre in Victoria, Australia. He holds an academic appointment as Adjunct Clinical Associate Professor at Monash University. Dixon is the Deputy Leader of the USANZ Genitourinary Oncology Special Advisory Group and the Urology Lead for the Victoria Complex Onco-Vascular Group (VCOG) at Austin Health. He completed his Doctor of Medical Science at the University of Melbourne in 2018, focusing on prostate cancer immunology, and subsequently undertook a prestigious Society of Urologic Oncology Fellowship at the University of Toronto and Princess Margaret Cancer Centre in 2019. Dixon's clinical and academic interests include robotic surgery, complex uro-oncology, prostate cancer biomarkers, and surgical equity. He has published over 50 peer-reviewed articles, secured more than $300,000 in research funding, and presented at major national and international conferences. He currently supervises four higher-degree research students and leads or collaborates on multiple clinical trials. He also serves as President of the Australian Chinese Medical Association Victoria (ACMAV), where he advocates for cultural inclusion, education, and health equity.

15th August 2025

Time	Session

15:30 17:00	Free Paper 【Podium】 (10) Oncology RCC (A) Cheryn SongKorea (Republic of) Moderator Assistance of the AI during RAPN - Surgical Navigation to Outcome PredictionArtificial Intelligence (AI) has deeply infiltrated many, if not every, aspects of our lives - both professional and ordinary corners. In medicine, it seems that any research dealing with large amounts of data has to employ AI in one way or another - from radiomics and radiogenomics to drug developments and simulations. In managing patients with renal tumors, majority of which now present with localized, small masses, I have focused on developing a tool to help establish a plan preoperatively and navigate throughout the surgery, in real-time with the help of the AI. Using kidney dynamic CT scans of 100 patients undergoing robot-assisted partial nephrectomy, software was developed to render a 3-dimensional image of the kidney harboring the tumor along with several other tools to enhance visualization of the lesion. Scans from an additional 30 patients were used to test for performance and validate the software. Before surgery, main renal pedicles, tumor location with respect to the major hilar vasculature, important landmarks including the calyces and branches of the vessels can be studied from all angles; tumor shape especially when it is not a perfect sphere can be visually presented and the excision of the tumor with desired margin thickness can be simulated to see what vital structures come into contact. Connected to the surgeon console through the TilePro® screen, it was designed to aid a urologic surgeon from before surgery as well as throughout the procedure as needed. In a phase I feasibility trial investigating the efficacy of the system, we confirmed that the integration of the 3D navigation system into RAPN was both feasible and safe, providing enhanced anatomical information while maintaining a consistent level of operative risk. The use of navigation system resulted in reduced renal parenchymal volume resected, suggesting potential benefits in renal function preservation. Subsequent phase 2 trial investigating accuracy of the reconstruction and a multicenter randomized controlled trial with surgeons with varying degrees of experience are underway to validate the benefits. On the other hand, the fundamental question as to best manage the small renal masses – i.e. indications for partial nephrectomies - still harbor some gray areas: is partial nephrectomy at all possible situations the best? In a given patient how will oncological and renal functional outcome differ between partial and radical nephrectomy? Previous statistical methods could only calculate risk ratios in the best-matched cohorts and the issue of counterfactual remained. Using demographic and tumor descriptive parameters of 1,448 patients with pT1N0M0 RCC patients from three academic centers, causal inference was modeled using a double machine learning algorithm to predict progression-free survival and eGFR at five years according to each surgical method: partial and radical nephrectomy. We identified significant factors and their contribution to the outcomes of interest to be included in the model. Selecting a model with the highest performance among various machine learning models for each outcome, an online interface was constructed. External validation confirmed the robustness, presenting AUROC of 0.758, AUPRC of 0.200 with XGBoost model for recurrence, and RMSE of 15.9 with GBM model for eGFR. By presenting numerical predictions of oncologic and functional outcomes associated with partial and radical nephrectomy in a given patient, our model facilitates individualized, evidence-based clinical decision-making. Chi-Ping HuangTaiwan Moderator Dixon WoonAustralia Moderator TICC - 1F 101D

Time

Session

15:30

17:00

Free Paper 【Podium】 (10)

Oncology RCC (A)

Cheryn SongKorea (Republic of) Moderator Assistance of the AI during RAPN - Surgical Navigation to Outcome PredictionArtificial Intelligence (AI) has deeply infiltrated many, if not every, aspects of our lives - both professional and ordinary corners. In medicine, it seems that any research dealing with large amounts of data has to employ AI in one way or another - from radiomics and radiogenomics to drug developments and simulations. In managing patients with renal tumors, majority of which now present with localized, small masses, I have focused on developing a tool to help establish a plan preoperatively and navigate throughout the surgery, in real-time with the help of the AI. Using kidney dynamic CT scans of 100 patients undergoing robot-assisted partial nephrectomy, software was developed to render a 3-dimensional image of the kidney harboring the tumor along with several other tools to enhance visualization of the lesion. Scans from an additional 30 patients were used to test for performance and validate the software. Before surgery, main renal pedicles, tumor location with respect to the major hilar vasculature, important landmarks including the calyces and branches of the vessels can be studied from all angles; tumor shape especially when it is not a perfect sphere can be visually presented and the excision of the tumor with desired margin thickness can be simulated to see what vital structures come into contact. Connected to the surgeon console through the TilePro® screen, it was designed to aid a urologic surgeon from before surgery as well as throughout the procedure as needed. In a phase I feasibility trial investigating the efficacy of the system, we confirmed that the integration of the 3D navigation system into RAPN was both feasible and safe, providing enhanced anatomical information while maintaining a consistent level of operative risk. The use of navigation system resulted in reduced renal parenchymal volume resected, suggesting potential benefits in renal function preservation. Subsequent phase 2 trial investigating accuracy of the reconstruction and a multicenter randomized controlled trial with surgeons with varying degrees of experience are underway to validate the benefits. On the other hand, the fundamental question as to best manage the small renal masses – i.e. indications for partial nephrectomies - still harbor some gray areas: is partial nephrectomy at all possible situations the best? In a given patient how will oncological and renal functional outcome differ between partial and radical nephrectomy? Previous statistical methods could only calculate risk ratios in the best-matched cohorts and the issue of counterfactual remained. Using demographic and tumor descriptive parameters of 1,448 patients with pT1N0M0 RCC patients from three academic centers, causal inference was modeled using a double machine learning algorithm to predict progression-free survival and eGFR at five years according to each surgical method: partial and radical nephrectomy. We identified significant factors and their contribution to the outcomes of interest to be included in the model. Selecting a model with the highest performance among various machine learning models for each outcome, an online interface was constructed. External validation confirmed the robustness, presenting AUROC of 0.758, AUPRC of 0.200 with XGBoost model for recurrence, and RMSE of 15.9 with GBM model for eGFR. By presenting numerical predictions of oncologic and functional outcomes associated with partial and radical nephrectomy in a given patient, our model facilitates individualized, evidence-based clinical decision-making.

Chi-Ping HuangTaiwan Moderator

Dixon WoonAustralia Moderator

TICC - 1F 101D